Andrew Nealen, TU Berlin, CG 11 Andrew Nealen TU Berlin Takeo Igarashi The University of Tokyo / PRESTO JST Olga Sorkine Marc Alexa TU Berlin Laplacian Mesh Optimization
Andrew Nealen, TU Berlin, CG 22 What is it ?
Andrew Nealen, TU Berlin, CG 33 Overview Motivation Problem formulation Laplacian mesh processing basics Laplacian mesh optimization framework Applications Triangle shape optimization Mesh smoothing Discussion
Andrew Nealen, TU Berlin, CG 44 Motivation Local detail preserving triangle optimization A Sketch-Based Interface for Detail Preserving Mesh Editing [Nealen et al. 2005]
Andrew Nealen, TU Berlin, CG 55 Motivation Local detail preserving triangle optimization A Sketch-Based Interface for Detail Preserving Mesh Editing [Nealen et al. 2005] Can we perform global optimization this way ? = L x
Andrew Nealen, TU Berlin, CG 66 Laplacian Mesh Processing Discrete Laplacians = Lx n cotangent : w ij = cot ij + cot ij uniform : w ij = 1
Andrew Nealen, TU Berlin, CG 77 Laplacian Mesh Processing Surface reconstruction n cotangent : w ij = cot ij + cot ij uniform : w ij = 1 = Lx LL y z x zz yy xx
Andrew Nealen, TU Berlin, CG 88 Laplacian Mesh Processing Surface reconstruction n zz yy xx y z x = L L Lc1c1 fix edit c2c2
Andrew Nealen, TU Berlin, CG 99 Laplacian Mesh Processing Least-squares solution n zz yy xx y z x = L L Lc1c1 fix edit c2c2 w1w1 w1w1 w2w2 w2w2 w Li Ax = b ATAATAx = bATAT (A T A) -1 x = bATAT Normal Equations
Andrew Nealen, TU Berlin, CG 10 Laplacian Mesh Processing Tangential smoothing n zz yy xx y z x = L L L fix c1c1 L L L
Andrew Nealen, TU Berlin, CG 11 L L L Laplacian Mesh Processing Tangential smoothing n zz yy xx y z x = fix c1c1
Andrew Nealen, TU Berlin, CG 12 L L L Laplacian Mesh Processing Tangential smoothing n zz yy xx y z x = fix c1c1
Andrew Nealen, TU Berlin, CG 13 More motivation… So: can we use such a system for global optimization ? =Lx
Andrew Nealen, TU Berlin, CG 14 Our Solution All vertices are (weighted) anchors Preserves global shape Uses existing LS framework Anchor + Laplacian weights determine result
Andrew Nealen, TU Berlin, CG 15 Framework Detail preserving tri shape optimization for L = L uni and f = cot (similar to local optimization) Mesh smoothing L = L cot (outer fairness) or L = L uni (outer and inner fairness) and f = 0 = Lxf WLWL WLWL p WPWP WPWP
Andrew Nealen, TU Berlin, CG 16 Tri Shape Optimization Detail preserving tri shape optimization = L uni x p WPWP WPWP
Andrew Nealen, TU Berlin, CG 17 Positional Weights
Andrew Nealen, TU Berlin, CG 18 Constant Weights
Andrew Nealen, TU Berlin, CG 19 Linear Weights
Andrew Nealen, TU Berlin, CG 20 CDF Weights
Andrew Nealen, TU Berlin, CG 21 CDF Weights
Andrew Nealen, TU Berlin, CG 22 Sharp Features
Andrew Nealen, TU Berlin, CG 23 Sharp Features
Andrew Nealen, TU Berlin, CG 24 Sharp Features
Andrew Nealen, TU Berlin, CG 25 Mesh Smoothing Mesh smoothing L = L cot (outer fairness) or L = L umb (outer and inner fairness) and f = 0 Controlled by W P and W L (Intensity, Features) Similar to Least-Squares Meshes [Sorkine et al. 04] = Lx0 WLWL WLWL p WPWP WPWP
Andrew Nealen, TU Berlin, CG 26 Using W P
Andrew Nealen, TU Berlin, CG 27 Using W P and W L
Andrew Nealen, TU Berlin, CG 28 Results
Andrew Nealen, TU Berlin, CG 29 Noisy
Andrew Nealen, TU Berlin, CG 30 Smoothed
Andrew Nealen, TU Berlin, CG 31 Original
Andrew Nealen, TU Berlin, CG 32 Tri Shape Optimization
Andrew Nealen, TU Berlin, CG 33 Smoothing Outer and Inner Fairness (L umb )
Andrew Nealen, TU Berlin, CG 34 Original
Andrew Nealen, TU Berlin, CG 35 Tri Shape Optimization
Andrew Nealen, TU Berlin, CG 36 Smoothing Outer Fairness only (L cot )
Andrew Nealen, TU Berlin, CG 37 Discussion The good... Easily controllable tri shape optimization and smoothing Leverages existing least squares framework Can replace tangential smoothing step for general remeshers ... and the not so good Euclidean distance is not Hausdorff distance, so error control is indirect Does rely on some (user) parameter tweaking
Andrew Nealen, TU Berlin, CG 38 Thank you ! Contact info Andrew Nealen Takeo Igarashi Olga Sorkine Marc Alexa